Skip to main content
Premium Trial:

Request an Annual Quote

Early Copy Number Changes Predict Progression of Barrett's Esophagus to Esophageal Cancer

NEW YORK – Copy number changes can predict esophageal cancer in patients with a precursor condition years before it would otherwise be detected, a new study has found.

As copy number alterations are often found in tumor tissue, but rarely in normal tissue, researchers led by the University of Cambridge's Rebecca Fitzgerald examined whether they could herald the development of esophageal adenocarcinoma among individuals with Barrett's esophagus. While Barrett's esophagus is a precursor condition to esophageal adenocarcinoma, not all patients progress to that stage — only about 0.3 percent of Barrett's esophagus patients develop esophageal adenocarcinoma each year.

By assessing copy number changes among Barrett's esophagus patients who progressed to esophageal adenocarcinoma and those who did not, the researchers developed a model to predict disease progression. As they reported on Monday in Nature Medicine, they found that these genomic changes could occur 10 years before histopathological transformation, indicating that genomic surveillance and classification could enable earlier treatment.

"This demonstrates that genomic risk stratification has a realistic potential to enable earlier intervention for high-risk conditions, and at the same time reduce the intensity of monitoring and even reduce overtreatment in cases of stable disease," Fitzgerald and her colleague wrote in their paper.

The researchers conducted shallow whole-genome sequencing — to an average depth of 0.4X — on a retrospective case-control cohort of 88 patients. More than 770 endoscopy samples had been collected from the patients during clinical surveillance of Barrett's esophagus. The researchers noted they used shallow whole-genome sequencing as it not only gave a genome-wide view of copy number changes but also had been optimized for use on formalin-fixed paraffin-embedded samples.

Overall, they found that samples from patients whose disease progressed to cancer exhibited generalized disorder across their genomes.

Based on the copy number data they generated and a measure of overall complexity, the researchers developed an elastic-net-regularized logistic regression model of progression and classification of disease. They validated the model in an independent cohort of 76 patients and orthogonally validated it using SNP array samples from 248 patients. 

The samples in their cohort with the highest relative risk were more than 20 times more likely to progress to esophageal cancer, while those with the lowest relative risk were 10 times less likely than average to have their disease progress. Using this, the researchers developed low, moderate, and high risk classifications, which they also applied to their validation cohort.

Slightly more than half the samples — 55 percent — from patients who did not progress were classified as low risk using the researchers' model. At the same time, 77 percent of the samples from patients who did progress were classified as high risk. 

Additionally, when the researchers assessed samples collected from patients throughout their disease history, they noted that most samples from patients who progressed were classified as high-risk over that entire timeframe while samples from patients who did not progress were consistently classified as low risk. About half the patients who progressed had samples deemed high risk from more than eight years before transformation.

When analyzed in conjunction with current Barrett's esophagus management guidelines, the researchers estimated that their approach would have led 54 percent of patients who progressed to receive earlier treatment. Meanwhile, they also estimated that of the patients who did not progress, 51 percent would have had less frequent endoscopies if their model had been applied.

The researchers cautioned, though, that this analysis relies on a relatively small cohort and that future studies with more longitudinal genomic data are needed to improve the sensitivity and specificity of their model.